Claude Opus 5.5 debuts matching Fable 5.1 at 40% lower cost, with best-ever alignment scores
sven_ai · x · 2026-09-28
Anthropic released Claude Opus 5.5, the first model in its Claude 5.5 family, performing at Claude Fable 5.1's level on most work while costing 40% less to run than Opus 5. Early testers completed a 680,000-line code migration in under a day; the model succeeded at cutting web app load times 39 of 40 times and topped a game-building comparison. It posts the best automated behavioral audit scores to date, is less prone to irreversible actions, and resists prompt injection better than Opus 5. The release was externally evaluated by Frontier Design and METR, and is Anthropic's first since calling for frontier pacing.
More from Models
- Anthropic Engineer's One-Prompt Opus 5.5 Supercut: Timestamped Demo Reel With Music — felixrieseberg · 2026-09-28
- 10 Wild Things People Are Building With Opus 5.5 — CodeByPoonam · 2026-09-28
- Opus 5.5 generates an 18-minute explainer video from prompts alone, no diffusion model, ~$5 in API calls — Afinetheorem · 2026-09-28
- Vals, backed by a16z, raises $40M to sell confidential task-based AI benchmarks — dl_weekly · 2026-09-28
- Merging Qwen3.6 & Qwen3.8 27B gives solid performance with fewer tokens — Danmoreng · 2026-09-28
- Purz pushes back: LLMs are not black holes — ComfyUI remains the connective tissue of genAI — PurzBeats · 2026-09-28